Impact of early vs. late initiation of continuous renal replacement therapy on composite outcomes including acute kidney disease and mortality: a multicenter propensity-matched cohort study (LINKA cohort)
Article information
Abstract
Background
The optimal timing of continuous renal replacement therapy (CRRT) initiation in acute kidney injury (AKI) remains uncertain, particularly regarding long-term kidney outcomes. This study evaluated whether early CRRT initiation improves the risk of acute kidney disease (AKD) or death.
Methods
In this multicenter retrospective cohort, 852 patients with baseline creatinine ≤4 mg/dL who received CRRT at eight tertiary hospitals were screened. Early initiation was defined as starting CRRT before KDIGO stage 3 or before severe oliguria (<0.3 mL/kg/hr for 24 hours). Propensity score matching (1:1) based on demographic, clinical, and laboratory variables yielded 746 matched patients. The primary outcome was a composite of AKD (≥50% increase in serum creatinine from baseline at 3 months) or death before 3-month follow-up. Multivariable logistic regression and stratified analyses by median baseline creatinine were performed.
Results
Early CRRT was associated with a significantly lower incidence of the composite outcome (odds ratio [OR], 0.40; 95% confidence interval [CI], 0.28–0.57; p < 0.001). The protective effect persisted in both the low-creatinine (OR, 0.46; 95% CI, 0.29–0.72) and high-creatinine subgroups (OR, 0.38; 95% CI, 0.21–0.68). A significant interaction between early CRRT and baseline creatinine (p = 0.04) suggested that the magnitude of benefit associated with early CRRT varied according to baseline renal function. Early CRRT was not significantly associated with 90-day mortality alone (adjusted OR, 0.69; p = 0.27).
Conclusion
Early CRRT initiation was associated with improved kidney-related outcomes, particularly in patients with lower baseline renal function. These findings support a more individualized approach to CRRT timing based on baseline kidney function.
Introduction
Acute kidney injury (AKI) is a common and severe complication in critically ill patients, and continuous renal replacement therapy (CRRT), also referred to as continuous kidney replacement therapy, is widely used for the management of severe acute kidney injury in critically ill patients. The timing of CRRT initiation has been a subject of ongoing debate [1]. While several prior studies have investigated early vs. late initiation of CRRT [2–5], results have been inconclusive, particularly regarding long-term renal outcomes and survival [6].
AKI is increasingly recognized as part of a spectrum of kidney dysfunction that may progress to acute kidney disease (AKD) [1], reflecting a subchronic phase of renal recovery or persistent kidney dysfunction. AKD has been associated with the progression to chronic kidney disease, dialysis dependence, and increased mortality [7–9]. Despite this, few studies have addressed how CRRT timing influences the development of AKD and subsequent outcomes.
Moreover, several AKI studies have suffered from survivor bias due to excluding patients who died before assessment of kidney recovery [10], thereby underestimating the benefit of early intervention. A composite endpoint combining AKD and mortality may provide a more comprehensive and clinically relevant measure of outcome [11].
Although large randomized controlled trials such as STARRT-AKI have evaluated accelerated renal replacement therapy (RRT) initiation strategies, these protocols were designed to test a standardized, early-threshold approach beginning at Kidney Disease: Improving Global Outcomes (KDIGO) stage 2. This differs from clinician-guided early CRRT in routine practice, which is typically initiated when patients show signs of impending renal or clinical deterioration. These conceptual differences highlight that existing trials and real-world practice address related but distinct questions, and they provide the rationale for further examining whether timely initiation before progression to severe AKI may influence the development of AKD.
In this multicenter retrospective cohort study, we aimed to investigate the impact of early vs. late CRRT initiation on a composite outcome of AKD and mortality. We hypothesized that early CRRT initiation would be associated with a reduced risk of adverse outcomes, particularly among patients with preserved baseline kidney function. To address confounding and selection bias, we used propensity score matching (PSM) and stratified analyses based on baseline serum creatinine.
Methods
Study design and population
We conducted a retrospective cohort study of adult patients (≥18 years) who received CRRT in the intensive care unit at eight tertiary hospitals between June 2016 and June 2022. A total of 1,474 patients who underwent CRRT were initially screened. Among these, patients with missing data on baseline creatinine or serum creatinine at 3 months (n = 335), or with baseline creatinine ≥4.0 mg/dL (n = 287), were excluded. After applying these exclusion criteria, 852 patients remained eligible for analysis. And 1:1 nearest-neighbor PSM without replacement was performed, yielding a final matched cohort of 746 patients (early CRRT, n = 373; late CRRT, n = 373) (Fig. 1).
Flowchart of patient selection in the LINKA multicenter CRRT cohort.
Flow diagram of study population selection. A total of 1,474 adult patients who received CRRT between June 2016 and June 2022 were screened. Patients with baseline serum creatinine (Cr) ≥4.0 mg/dL, missing baseline Cr values, or missing 3-month Cr follow-up (FU) values were excluded. After applying these criteria, 852 patients remained eligible. Propensity score matching (1:1 nearest-neighbor without replacement) based on demographic, clinical, and laboratory covariates yielded 746 matched patients (early initiation, n = 373; late initiation, n = 373) included in the final analytic cohort. Early CRRT was defined as initiation of CRRT before progression to Kidney Disease: Improving Global Outcomes (KDIGO) stage 3 acute kidney injury (urine output, <0.3 mL/kg/hr for 24 hours). Late CRRT was defined as initiation of CRRT at KDIGO stage 3 acute kidney injury.
CRRT, continuous renal replacement therapy; LINKA, LINKing health medical records, biospecimens, and biosignals in Korean patients with Acute kidney injury.
Definition of continuous renal replacement therapy timing
Early and late CRRT initiation were defined using objective, clinically meaningful criteria suitable for retrospective analysis. Early CRRT was defined as initiation before the development of KDIGO stage 3 criteria, reflecting a clinically recognized point of impending deterioration at which CRRT is often considered in practice. Late initiation was defined as CRRT started after KDIGO stage 3 criteria were met. These definitions were chosen to align with real-world decision-making and to ensure reproducibility in a multicenter retrospective design.
Baseline variables and outcomes
The primary outcome was a composite of AKD and all-cause mortality within 90 days, to account for mortality as a competing risk for renal recovery following severe AKI. Baseline serum creatinine was defined as the lowest creatinine value measured within 1 year prior to hospitalization. If no preadmission value was available, the lowest creatinine measured during the initial admission period was used as the baseline value. Patients with end-stage kidney disease or with baseline creatinine ≥4.0 mg/dL were excluded from the analysis. This baseline creatinine value was used consistently as the reference for KDIGO AKI staging and for AKD determination at 3 months. AKD was defined based on KDIGO criteria as a ≥50% increase in serum creatinine from baseline at 3 months. Serum creatinine at 3 months was defined as the value obtained closest to day 90. Because follow-up creatinine measurements did not occur on a single uniform date across centers, we used the creatinine value within the observed clinical follow-up distribution. Among included patients, the interquartile range (25th–75th percentile) of the measurement window was 89–103 days, and values within this range were used for the 3-month assessment. Mortality data were obtained from hospital records and national death registry linkage.
Baseline characteristics included demographic variables (age, sex), comorbidities (hypertension, diabetes), severity-related variables (mechanical ventilation, vasopressor use, Glasgow Coma Scale [GCS] score), laboratory data (hemoglobin, platelet count, international normalized ratio [INR], potassium), and AKI etiology (septic, ischemic, nephrotoxic, postoperative, others) (Table 1).
Clinical characteristics of patients before and after propensity score matching
Continuous renal replacement therapy practice
This multicenter study included patients treated at eight tertiary referral hospitals. Although CRRT initiation and management were guided by local institutional protocols, all centers followed standard contemporary CRRT practices. CRRT was typically initiated by nephrologists or intensivists based on overall clinical assessment. Continuous venovenous hemodiafiltration was the predominant modality, using commercially available CRRT machines (e.g., Prismaflex/Prismax [Baxter] or MultiFiltrate [Fresenius Medical Care]). Blood flow rates and effluent doses were prescribed within commonly accepted clinical ranges. To minimize center-related variability, the definitions of early and late CRRT initiation were standardized using KDIGO-based criteria and applied uniformly across all institutions.
Statistical analysis
To reduce confounding and balance baseline characteristics between groups, 1:1 nearest-neighbor PSM without a caliper was performed using a logistic regression model incorporating the following covariates: age, sex, hypertension, diabetes, five AKI etiologies, mechanical ventilation, GCS score, BMI, hemoglobin, platelet count, INR, potassium, vasopressor use, CRRT duration, and baseline serum creatinine. Matching was done using nearest-neighbor matching without replacement.
Logistic regression was used to assess the association between early vs. late CRRT initiation and the primary composite outcome. Adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were reported. Subgroup analysis was performed stratified by baseline serum creatinine (low vs. high, based on median split). Interaction testing was also conducted to assess effect modification by baseline creatinine.
All analyses were conducted using R version 4.4.2 (R Foundation for Statistical Computing). A two-sided p-value <0.05 was considered statistically significant.
Results
A total of 852 patients who received CRRT were screened. Patients in the early and late CRRT groups differed in several demographic and clinical characteristics, including age, AKI etiology, and laboratory values. After PSM, 373 matched pairs were generated with adequate covariate balance, with most standardized mean differences below 0.1. Among patients in the early CRRT group, 77 patients (20.6%) met KDIGO stage 1 criteria and 104 patients (27.9%) met KDIGO stage 2 criteria based on serum creatinine criteria at the time of CRRT initiation. The remaining patients did not meet creatinine-based KDIGO stage 1 or 2 criteria but initiated CRRT prior to progression to stage 3. These patients likely met urine output–based AKI criteria. The two groups showed comparable distributions of comorbidities, hemodynamic status, and kidney function parameters. Baseline characteristics included demographic variables (age, sex, BMI), comorbidities (hypertension, diabetes), severity-related variables (mechanical ventilation, vasopressor use, GCS score, CRRT duration), laboratory data (hemoglobin, platelet count, INR, potassium), and AKI etiology (septic, ischemic, nephrotoxic, postoperative, others) (Table 1). At the time of CRRT initiation, serum creatinine levels were lower in the early CRRT group compared with the late CRRT group (mean ± standard deviation, 2.01 ± 0.88 mg/dL vs. 3.40 ± 1.93 mg/dL).
In the propensity score-matched cohort (n = 746), crude clinical outcomes are summarized in Table 2. The incidence of the composite outcome was lower in the early CRRT group than in the late CRRT group. AKD at 3 months occurred less frequently in the early CRRT group. Death within 90 days occurred in 18 patients (4.8%) in the early CRRT group and 25 patients (6.7%) in the late CRRT group. In adjusted analyses, early CRRT initiation was not significantly associated with 90-day mortality. Early CRRT initiation was statistically significantly associated with a lower risk of the composite outcome (adjusted OR, 0.37; 95% CI, 0.27–0.50; p < 0.001) compared with late initiation (Fig. 2).
Crude clinical outcomes according to CRRT timing
Association between early CRRT initiation and the primary composite outcome.
Adjusted odds ratios (ORs) for the primary composite outcome of acute kidney disease or 90-day mortality comparing early vs. late CRRT initiation in the propensity-score matched cohort (n = 746). Early CRRT was associated with a significantly lower risk of the composite outcome (adjusted OR, 0.37; 95% confidence interval, 0.27–0.50; p < 0.001). Multivariable adjustment included age, sex, comorbidities, body mass index (BMI), acute kidney injury etiology (septic, postoperative, ischemic, nephrotoxic, others), severity indicators (Glasgow Coma Scale [GCS] score, mechanical ventilation, vasopressor use, CRRT duration), and laboratory parameters (hemoglobin, platelets, international normalized ratio [INR], potassium, and baseline creatinine). For continuous variables (age, BMI, hemoglobin, baseline serum creatinine, potassium, and CRRT duration), ORs represent the change in risk associated with a one-unit increase in each variable. Accordingly, ORs <1 indicate a lower risk with higher values of the variable, whereas ORs >1 indicate a higher risk with increasing values. The dashed vertical line indicates an OR of 1.0.
CRRT, continuous renal replacement therapy; DM, diabetes mellitus; INR, international normalized ratio.
This association remained significant after adjustment for age, sex, comorbidities (hypertension, diabetes mellitus), AKI etiology, surgical context, baseline illness severity (GCS score, BMI, mechanical ventilation, vasopressor use), and baseline laboratory values (hemoglobin, platelets, INR, potassium, creatinine). In the multivariable logistic regression analysis for AKD as an independent outcome, early CRRT initiation was significantly associated with a reduced risk of AKD (OR, 0.37; 95% CI, 0.26–0.53; p < 0.001). However, when mortality was analyzed separately, early CRRT did not significantly reduce the risk of 90-day mortality alone (adjusted OR, 0.69; 95% CI, 0.35–1.33; p = 0.27).
Among covariates, several factors showed statistically significant associations with the composite outcome. Higher baseline hemoglobin (OR, 0.90; 95% CI, 0.84–0.98; p = 0.01) and higher body mass index (OR, 0.95; 95% CI, 0.91–0.98; p = 0.006) and baseline serum creatinine (OR, 0.31; 95% CI, 0.23–0.40; p < 0.001) were significantly associated with the composite outcome.
Stratified analysis by baseline creatinine
To explore the heterogeneity of the effect according to baseline renal function, we conducted a stratified analysis using the median baseline creatinine value of 1.01 mg/dL. Patients were divided into low baseline creatinine (<1.01 mg/dL) and high baseline creatinine (≥1.01 mg/dL) subgroups.
In both strata, early CRRT initiation remained significantly associated with a reduced risk of the composite outcome. In the low creatinine group, early initiation was associated with 54% lower odds of composite outcome (OR, 0.46; 95% CI, 0.29–0.72; p = 0.0007), whereas in the high creatinine group, the effect size was similar (OR, 0.38; 95% CI, 0.21–0.68; p = 0.001) (Fig. 3).
Stratified analysis according to baseline serum creatinine (Cr) levels.
Stratified analysis of early vs. late continuous renal replacement therapy (CRRT) initiation according to baseline serum Cr levels, divided at the median value of 1.01 mg/dL. Early CRRT showed significant benefit in both subgroups: low baseline Cr (<1.01 mg/dL): OR, 0.46 (95% CI, 0.29–0.72; p = 0.0007); high baseline Cr (≥1.01 mg/dL): OR, 0.38 (95% CI, 0.21–0.68; p = 0.001). These findings indicate that early CRRT confers consistent benefit across different degrees of baseline renal function, supporting the robustness of the primary results.
CI, confidence interval; OR, odds ratio.
Interaction analysis
To evaluate whether the effect of early CRRT initiation varied according to baseline renal function, we included an interaction term between early CRRT and baseline serum creatinine in the multivariable logistic regression model. The interaction term was statistically significant (interaction coefficient = –0.60, p = 0.04), indicating a modifying effect of baseline creatinine on the association between early CRRT and the composite outcome. In this interaction model, early CRRT as an independent effect was not significantly associated with the composite outcome (adjusted OR, 0.72; p = 0.27) (Fig. 4).
Interaction between early CRRT initiation and baseline creatinine (Cr).
Multivariable logistic regression model including an interaction term between early CRRT initiation and baseline serum Cr. The interaction term was statistically significant (interaction coefficient, −0.60, p = 0.04), indicating effect modification by baseline renal function. Early CRRT was not significant as a main effect in the interaction model (adjusted OR, 0.72; p = 0.27), suggesting that the magnitude of benefit depends on baseline Cr rather than being uniform across all patients. Greater benefit was observed in those with lower baseline renal reserve.
CI, confidence interval; CRRT, continuous renal replacement therapy; OR, odds ratio.
Fig. 5 illustrates the predicted probability of the composite outcome across the continuum of baseline serum creatinine, derived from the multivariable logistic regression model including the interaction term between early CRRT initiation and baseline creatinine. The curves depict the estimated risk for early and late CRRT initiation, with shaded areas representing 95% CIs. These predicted risks were derived from the same multivariable model including the interaction term.
Interaction effect between early CRRT initiation and baseline creatinine on predicted risk of the composite outcome.
Predicted probability curves illustrating the interaction between early CRRT initiation and baseline serum creatinine on the risk of the primary composite outcome (acute kidney disease or 90-day mortality). The solid orange line represents the predicted risk in the early CRRT group, and the dashed red line represents the predicted risk in the late CRRT group across a continuum of baseline creatinine levels. Early CRRT is associated with a consistently lower predicted risk than late CRRT, with the absolute risk difference (shaded gray area) becoming more pronounced as baseline creatinine rises from normal to moderately elevated levels and then attenuating at the highest creatinine values. This pattern illustrates the statistically significant interaction, suggesting that the renoprotective effect of early CRRT is more evident in patients with relatively higher baseline serum creatinine within the studied range.
CRRT, continuous renal replacement therapy.
Persistent renal replacement therapy dependence
Persistent RRT dependence at the last available follow-up occurred in 66 of 373 patients (17.7%) in the early CRRT group and 95 of 373 patients (25.5%) in the late CRRT group. In multivariable logistic regression, early CRRT initiation was associated with a lower likelihood of ongoing dialysis requirement (OR, 0.55; 95% CI, 0.38–0.80; p = 0.002). Hemoglobin level (OR, 0.91; 95% CI, 0.84–0.99; p = 0.03) was also a significant predictor of dialysis dependence. Diabetes mellitus (adjusted OR, 1.60; 95% CI, 1.08–2.39; p = 0.02), higher baseline creatinine (OR, 1.27; 95% CI, 1.04–1.55; p = 0.02), and longer CRRT duration (OR, 1.02; 95% CI, 1.01–1.03; p = 0.002) were associated with increased risk of RRT dependence. These findings were directionally consistent with the forest plot (Fig. 6).
Adjusted ORs for RRT dependence at the last available follow-up.
Forest plot showing adjusted ORs and 95% CIs for RRT dependence at the last available follow-up, derived from a multivariable logistic regression model including early CRRT timing and baseline covariates (age, sex, BMI, diabetes mellitus [DM], hypertension, acute kidney injury etiology, mechanical ventilation, Glasgow Coma Scale (GCS) score, hemoglobin, platelet count, INR, potassium, vasopressor use, baseline creatinine, and CRRT duration). Early CRRT was associated with significantly lower odds of RRT dependence (OR, 0.55; 95% CI, 0.38–0.80; p = 0.002). DM was associated with higher RRT dependence (OR, 1.60; 95% CI, 1.08–2.39; p = 0.02), whereas higher hemoglobin was protective (OR, 0.91; 95% CI, 0.84–0.99; p = 0.03). Higher baseline creatinine (OR, 1.27; 95% CI, 1.04–1.55; p = 0.02) and longer CRRT duration (OR, 1.02; 95% CI, 1.01–1.03; p = 0.003) were associated with increased risk of RRT dependence.
BMI, body mass index; CI, confidence interval; CRRT, continuous renal replacement therapy; INR, international normalized ratio; OR, odds ratio; RRT, renal replacement therapy.
Discussion
In this multicenter propensity score-matched cohort study, early initiation of CRRT was associated with a significant reduction in the risk of composite adverse outcomes, which included AKD progression and 90-day mortality. Notably, early CRRT was significantly associated with a lower risk of AKD. In contrast to kidney-related outcomes, early CRRT was not associated with a statistically significant reduction in short-term mortality, which is consistent with prior randomized trials evaluating RRT timing [4]. Moreover, in a separate analysis, early CRRT initiation was also independently associated with a lower risk of subsequent dialysis dependence, which represents a key component of major adverse kidney events. Together, these findings indicate that early CRRT may influence both short-term renal recovery and longer-term dialysis dependency trajectories. In addition to creatinine-based AKD, we assessed persistent RRT dependence as a complementary measure of renal recovery. Early CRRT initiation was associated with a significantly lower likelihood of ongoing dialysis requirement at follow-up (adjusted OR, 0.55), suggesting that its renoprotective effect extends beyond biochemical improvement alone. Although follow-up timing varied across centers, the consistent direction of benefit supports a broader recovery advantage with earlier intervention. Interestingly, higher baseline serum creatinine was inversely associated with the composite outcome (OR, 0.31; 95% CI, 0.23–0.40; p < 0.001), a finding that may appear paradoxical given the expected link between renal dysfunction and poor prognosis. This inverse relationship between baseline creatinine and the composite outcome is likely due to the definition of AKD, which is based on relative increases in serum creatinine. Patients with lower baseline values require smaller absolute changes in creatinine to meet the AKD definition, whereas those with higher baseline values may not fulfill the criterion despite similar or greater absolute increases. This structural bias could lead to an overrepresentation of AKD among patients with initially lower creatinine values, contributing to the apparent protective effect of higher baseline creatinine in the model.
To further explore whether the effect of early CRRT differed according to the severity of renal dysfunction at the time of initiation, we conducted both stratified analyses and an interaction term analysis based on baseline serum creatinine. Our stratified analyses confirmed that early CRRT initiation conferred a protective effect in both patients with preserved and impaired baseline renal function, albeit with varying effect magnitudes. Importantly, interaction term analysis confirmed a statistically significant effect modification by baseline creatinine, indicating that the efficacy of early CRRT is not uniform across all levels of kidney injury. The predicted-risk curves further support the presence of effect modification by baseline renal function. As shown in Fig. 5, the separation between early and late CRRT initiation varies across baseline creatinine levels, suggesting that the magnitude of benefit associated with early CRRT is not constant but depends on the degree of renal dysfunction at presentation. Importantly, these findings should be interpreted as exploratory and hypothesis-generating, given the observational nature of the analysis and the use of creatinine-based definitions. Notably, early CRRT was associated with a reduced risk of adverse outcomes across a broad range of baseline creatinine levels, although the magnitude of benefit varied according to baseline renal function. This suggests that preserved kidney function at presentation may not necessarily warrant delaying early intervention, especially when other indications are present. In the multivariable model including the interaction term, early CRRT alone was not statistically significant, but the interaction between early CRRT and baseline serum creatinine was. The interaction analysis further supports that while benefit is observed across the board, it may be more consistently detectable in those with reduced renal reserve, who are at higher risk of AKD progression. This pattern is visually demonstrated in the predicted-risk curves (Fig. 5), where the separation between early and late CRRT widens as baseline creatinine increases and then gradually narrows again at the highest creatinine values.
These results underscore the importance of considering individual renal context in timing decisions and suggest that in patients with more impaired renal function at presentation, early intervention may be essential to prevent the subacute progression of kidney injury. This implies that delayed therapy in these patients may miss a critical window for renal recovery.
This study has several limitations. First, as a retrospective observational study, it is subject to residual confounding despite adjustment through PSM and multivariable modeling. Second, although we accounted for major clinical and biochemical variables, important unmeasured factors—such as nephrotoxin exposure, detailed fluid balance, and center- or physician-level practice patterns—may have influenced both the decision to initiate CRRT and subsequent outcomes. In particular, urgent indications for CRRT initiation (e.g., refractory hyperkalemia, severe acidosis, or pulmonary edema) could not be evaluated consistently across centers due to documentation variability. The inability to incorporate these urgent triggers represents an inherent limitation of retrospective data and was one reason for adopting objectively extractable KDIGO-based criteria to define early and late initiation. Third, the assessment of renal outcomes was based primarily on serum creatinine and 24-hour urine output, which may not fully reflect underlying kidney function or distinguish between transient and sustained injury. Direct measurements of glomerular filtration rate (GFR) or renal biomarkers were not available, and estimated GFR could not be reliably assessed in this retrospective multicenter cohort due to incomplete data. Consequently, baseline renal function was evaluated using serum creatinine, which may be influenced by sarcopenia in critically ill patients. To partially address this limitation, we conducted stratified and interaction analyses based on baseline creatinine as exploratory assessments of potential effect modification. Additionally, formal severity indices such as the APACHE-II (Acute Physiology and Chronic Health Evaluation II) or the SOFA (Sequential Organ Failure Assessment) or a comprehensive comorbidity index like Charlson Comorbidity Index could not be incorporated because several required physiological components (e.g., detailed respiratory parameters, bilirubin, and neurologic assessments at standardized time points) and identified cardiovascular comorbidities were not consistently available across centers. Instead, we adjusted for key surrogate indicators of severity, including vasopressor use, mechanical ventilation, GCS score, and major laboratory parameters. Therefore, prospective studies incorporating detailed renal functional assessments, such as measured GFR, biomarker profiles, and standardized timing criteria, are warranted to more accurately define the causal impact of early CRRT initiation and validate these findings in diverse clinical settings.
Despite these limitations, the findings provide important clinical implications. First, our findings may suggest a possible renoprotective effect of early CRRT initiation, although no clear benefit in short-term mortality was observed. Second, they highlight the need to consider baseline renal function when evaluating the timing of CRRT, as its benefit appears modifiable by underlying kidney status. Third, our findings support current clinical practice suggesting that early CRRT can be considered even in patients with relatively preserved serum creatinine levels when clear non-renal indications are present.
In conclusion, this multicenter propensity score-matched cohort study suggests that early initiation of CRRT is associated with improved kidney-related outcomes, including a lower risk of AKD and reduced dialysis dependence. The observed association across a broad range of baseline renal function highlights the importance of individualized decision-making when considering the timing of CRRT initiation. While these findings should be interpreted cautiously given the observational design, they support current clinical practice in which early CRRT may be considered when clinically indicated, even in patients with relatively preserved serum creatinine levels. Further prospective studies with standardized timing criteria, comprehensive severity assessment, and longer-term follow-up are needed to clarify the causal relationship between CRRT timing and renal recovery.
Notes
Conflicts of interest
All authors have no conflicts of interest to declare.
Funding
This study was supported by the National Institute of Health research project (No. 2023-ER1105-00), a cooperative research fund from the Korean Society of Nephrology (2023), and in part by Seoul National University Bundang Hospital (grant No. 18-2018-0031).
Data sharing statement
The data presented in this study are available from the corresponding author upon reasonable request.
Authors’ contributions
Conceptualization: SM, SK
Data curation, Investigation, Supervision: JNA
Formal analysis: JNA, SM
Funding acquisition, Project administration, Resources: SGK
Methodology, Software, Validation: SK
Visualization: SM
Writing–original draft: JNA, SGK
Writing–review & editing: All authors
All authors read and approved the final manuscript.